K. Gaffour, M. K. Benhaoua, Somdip Dey, A. Singh, Abou El hassen Benyamina
{"title":"Dynamic clustering approach for run-time applications mapping on NoC-based multi/many-core systems","authors":"K. Gaffour, M. K. Benhaoua, Somdip Dey, A. Singh, Abou El hassen Benyamina","doi":"10.1109/EDiS49545.2020.9296439","DOIUrl":null,"url":null,"abstract":"Networks-on-Chips have been suggested as a powerful solution to communication problems in multi/many-core architectures. The communications mapping consists of one of the crucial steps in architectural exploration in NoC-based multi/many-core architectures. Most works that are reported in the literature use virtual static clustering (the platform is devised in a different region with the same size). Parallelized mapping of communications tasks of the same application onto different regions causes degradation of system performance, especially when the number of tasks is greater than the number of processors of the chosen cluster. In this paper, a new dynamic clustering mapping strategy is proposed to place all the tasks of the same application in the same region. We have applied our proposed approach under different heuristics of applications tasks mapping. Comparative study of our proposed strategy provides significant performance improvements over the static clustering methodology.","PeriodicalId":119426,"journal":{"name":"2020 Second International Conference on Embedded & Distributed Systems (EDiS)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"5","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 Second International Conference on Embedded & Distributed Systems (EDiS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/EDiS49545.2020.9296439","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 5
Abstract
Networks-on-Chips have been suggested as a powerful solution to communication problems in multi/many-core architectures. The communications mapping consists of one of the crucial steps in architectural exploration in NoC-based multi/many-core architectures. Most works that are reported in the literature use virtual static clustering (the platform is devised in a different region with the same size). Parallelized mapping of communications tasks of the same application onto different regions causes degradation of system performance, especially when the number of tasks is greater than the number of processors of the chosen cluster. In this paper, a new dynamic clustering mapping strategy is proposed to place all the tasks of the same application in the same region. We have applied our proposed approach under different heuristics of applications tasks mapping. Comparative study of our proposed strategy provides significant performance improvements over the static clustering methodology.